No evidence for contextual cueing beyond explicit recognition

Sascha Meyen, Miguel A. Vadillo, Ulrike von Luxburg, Volker H. FranzPsychonomic 2023 · DOI

TL;DR

This paper argues that the common conclusion of "implicit recognition" in contextual cueing tasks is flawed. By reanalyzing 20 studies, the authors show that the evidence for a superior "implicit" sensitivity in response times, compared to explicit reports, is largely absent when analyzed correctly. This suggests that observed effects might be better explained by weak explicit recognition rather than a separate, unconscious process.

What Problem Does This Paper Address?

The field of cognitive psychology often investigates how we learn and process information without conscious awareness, a phenomenon termed "implicit" or "unconscious" processing. A prominent paradigm for studying this is contextual cueing, where participants search for a target among distractors. Researchers observe that participants become faster at finding the target when the spatial arrangement of distractors (the "context") is repeated across trials. This speed advantage is often interpreted as evidence of learning the spatial regularities.

The standard reasoning to claim implicit learning relies on two observations:

  1. A reliable response time (RT) difference: Participants are faster on repeated trials compared to novel ones.
  2. A close-to-chance explicit recognition performance: When asked directly if they remember a specific configuration, participants perform poorly, suggesting they don't consciously recall it.

From this pattern, researchers infer an "Indirect Task Advantage" (ITA), where the RT measure shows higher sensitivity than explicit reports, leading to the conclusion of "implicit recognition." However, this paper argues this reasoning is flawed because it doesn't properly compare the sensitivities of the two measures.

Methods

The core of the paper's contribution is the introduction and application of a sensitivity comparison method. Instead of just comparing RT differences to explicit report accuracy, the authors propose a more rigorous approach:

  • Direct Sensitivity (d'direct): This is calculated from participants' explicit recognition task performance (hit rates and false alarm rates), using signal detection theory. This is a direct measure of conscious recognition.
  • Indirect Sensitivity (d'indirect): This is estimated from the search task RT data. Crucially, they use an ideal observer analysis. This assumes a participant has full knowledge of their RTs and uses them optimally to predict whether a configuration was repeated or new. This simulates a situation where implicit recognition might be driving RTs.
  • Sensitivity Comparison: The key step is to directly compare d'direct and d'indirect. An Indirect Task Advantage (ITA) is only established if d'indirect is significantly greater than d'direct.

The authors then apply this method in a reanalysis of 20 published contextual cueing studies. They also develop a method to estimate d'indirect from reported statistics (like t-tests) when full trial-by-trial data isn't available. They employ a "benefit-of-the-doubt" approach, biasing their analysis to favor finding an ITA, to make their conclusions more robust.

Datasets

The paper reanalyzes data from 20 previously published contextual cueing studies. For some of these studies, they obtained full trial-by-trial data from the original authors. For others, they used reported statistics. No new datasets were created or introduced.

Key Insights and Ideas

  • The "Standard Reasoning" is Flawed: The common practice of inferring implicit recognition from a combination of faster RTs and poor explicit recall is methodologically unsound. It doesn't directly compare the sensitivity of the measures.
  • Sensitivity Comparison is Crucial: To claim an "Indirect Task Advantage" (ITA), one must demonstrate that the sensitivity of the indirect measure (RTs) is genuinely higher than the sensitivity of the direct measure (explicit reports).
  • Lack of Consistent Evidence for ITAs: Across a reanalysis of 20 studies, the authors found very little consistent evidence for ITAs, even when using a generous "benefit-of-the-doubt" approach. This suggests that the observed RT effects might not reflect a separate, unconscious recognition process.
  • Weak Explicit Recognition is a Plausible Explanation: The absence of ITAs implies that the observed RT effects could be fully explained by weak, but present, explicit recognition. This challenges the need for a separate "implicit recognition" construct.
  • Correlation-Based Arguments are Also Problematic: The paper also critiques arguments based on low correlations between RT effects and explicit recognition. These low correlations can arise from factors like measurement unreliability or non-linear relationships, rather than true independence of processes.

Key Results

  • In a reanalysis of 20 contextual cueing studies, only a small fraction (37 out of 56 conditions/experiments) showed a numerical difference favoring an ITA.
  • When considering statistical significance and measurement error, the vast majority of studies (19 out of 20) did not provide consistent evidence for an ITA.
  • Even in the one study that showed some evidence for ITAs (Conci & von Mühlenen, 2011), the authors argue these effects occurred before participants could have learned the display configurations, suggesting they stemmed from other factors.
  • Bayesian analyses also tended to favor the null hypothesis (no ITA) in most cases.
  • Individual-level analysis of a large dataset (Colagiuri & Livesey, 2016) showed that only 6.3% of participants demonstrated a significant ITA, a proportion consistent with chance.

Limitations and Future Work

  • Measurement Error: The reanalysis based on reported statistics incurs more measurement error than using full trial-by-trial data. However, the authors argue their "benefit-of-the-doubt" approach (using a higher variance ratio estimate) compensates for this.
  • Neuro-muscular Noise: RTs are subject to neuro-muscular noise, which can reduce observed sensitivity. The authors suggest that proponents of implicit recognition need to provide evidence for ITAs after accounting for this noise.
  • Binary vs. Non-Binary Recognition Tasks: The current sensitivity comparison method is primarily applicable to binary explicit recognition tasks. Future work could extend this to more complex, non-binary tasks (like quadrant generation tasks).
  • Learned Displays: The paper acknowledges that contextual cueing effects might stem from only a subset of learned displays. However, selecting these displays post-hoc can introduce regression-to-the-mean issues, making it difficult to draw firm conclusions.
  • Meta-Analysis: The authors abstained from a meta-analysis due to the significant heterogeneity in study designs, populations, and research questions. They argue that each study claiming implicit recognition should provide self-contained evidence.
  • Future Research Directions: The authors suggest investigating ITAs in specific subgroups (e.g., by age), establishing double dissociations, and developing methods to identify learned displays without introducing methodological biases. They also call for more rigorous statistical approaches to avoid misinterpretations.

Summary updated

Abstract

Many studies claim that visual regularities can be learned unconsciously and without explicit awareness. For example in the contextual cueing paradigm, studies often make claims using a standard reasoning based on two results: (1) a reliable response time (RT) difference between repeated vs. new stimulus displays and (2) a close-to-chance sensitivity when participants are asked to explicitly recognize repeated stimulus displays. From this pattern of results, studies routinely conclude that the sensitivity of RT responses is higher than that of explicit responses-an empirical situation we call Indirect Task Advantage (ITA). Many studies further infer from an ITA that RT effects were driven by a form of recognition that exceeds explicit memory: implicit recognition. However, this reasoning is flawed because the sensitivity underlying RT effects is never computed. To properly establish a difference, a sensitivity comparison is required. We apply this sensitivity comparison in a reanalysis of 20 contextual cueing studies showing that not a single study provides consistent evidence for ITAs. Responding to recent correlation-based arguments, we also demonstrate the absence of evidence for ITAs at the level of individual participants. This lack of ITAs has serious consequences for the field: If RT effects can be fully explained by weak but above-chance explicit recognition sensitivity, what is the empirical content of the label "implicit"? Thus, theoretical discussions in this paradigm-and likely in other paradigms using this standard reasoning-require serious reassessment because the current data from contextual cueing studies is insufficient to consider recognition as implicit.
BibTeX
@inproceedings{meyen2023no,
  author = {Sascha Meyen and Miguel A. Vadillo and Ulrike von Luxburg and Volker H. Franz},
  title = {{No evidence for contextual cueing beyond explicit recognition}},
  booktitle = {Psychonomic},
  year = {2023},
  doi = {10.3758/s13423-023-02358-3},
}
Figure